# Zalando Lounge Awin Feed Analysis

**Date:** 2026-02-23
**Feed:** Custom URL with 92 columns (expanded from default)
**File:** `D:\sneakerpicks-feeds\zalando-lounge-custom.csv.gz` (54 MB compressed)

## TL;DR

**❌ The custom feed URL does NOT solve the matching problem.** Despite requesting 92 columns including EAN, GTIN, UPC, model_number, mpn, product_model, and custom_1-9, Zalando Lounge provides **ZERO** data in any of these fields. The product names remain completely generic ("Sneakers laag"). This is a Zalando Lounge limitation — they simply don't expose product identifiers through the Awin feed.

## Key Numbers

| Metric | Value |
|--------|-------|
| Total products | 216,989 |
| Shoe products (broad) | 42,327 |
| Sneaker products | 13,882 |
| Products with EAN | **0 (0%)** |
| Products with product_GTIN | **0 (0%)** |
| Products with UPC | **0 (0%)** |
| Products with model_number | **0** |
| Products with mpn | **0** |
| Products with product_model | **0** |
| custom_1 through custom_9 | **All 0 (0%)** |
| Fashion:category | **Empty for all** |
| Fashion:suitable_for | Filled (male/female/unisex) |
| merchant_product_id | 216,989 (100%) — but opaque codes like `07915P004-K11` |

## Sneaker Product Names — Still Generic

Only **9 unique product names** across all 13,882 sneaker products:

| Name | Count |
|------|-------|
| Sneakers laag | 12,262 |
| Sneakers hoog | 1,228 |
| Trainingsschoen | 253 |
| Skateschoenen | 39 |
| Handbalschoenen | 33 |
| Volleybalschoenen | 30 |
| Instappers | 28 |
| Babyschoenen | 7 |
| Tennisschoenen voor alle ondergronden | 2 |

**This is identical to the previous feed issue** — generic Dutch category names, no actual model names.

## Brand Distribution (Shoes, Top 20)

Good sneaker brands are present but products can't be matched to models:

| Brand | Count |
|-------|-------|
| Puma | 1,581 |
| New Balance | 1,335 |
| Converse | 1,003 |
| Nike Sportswear | 828 |
| adidas performance | 756 |
| adidas Originals | 750 |
| Nike Performance | 656 |
| Vans | 590 |
| Tommy Hilfiger | 548 |
| Polo Ralph Lauren | 709 |

## What's Usable

The only semi-useful identifier is `merchant_product_id` (e.g., `07915P004-K11`). These are Zalando's internal SKU codes. They are:
- **Opaque** — no model name encoded
- **Unique per product** — could theoretically be used to scrape Zalando Lounge pages for actual product details
- **Not standardized** — can't be cross-referenced with other feeds

## Comparison with Previous Analysis

| Aspect | Previous Feed | This Custom Feed |
|--------|---------------|-----------------|
| Total products | ~8,403 | 216,989 |
| Columns | ~20 | 92 |
| EAN/GTIN | 0 | 0 |
| Model names | Generic | Generic |
| custom_1-9 | Not requested | All empty |
| Fashion:category | Not requested | All empty |

The feed is **much larger** (217K vs 8.4K) — likely because the previous analysis only looked at a subset. But the data quality problem is identical.

## Recommendation

**The Awin feed for Zalando Lounge is a dead end for product matching.** Zalando Lounge deliberately strips product identifiers from their affiliate feed data.

### Possible alternatives:
1. **Scraping Zalando Lounge product pages** — The `merchant_deep_link` URLs lead to actual product pages which likely contain the real product name and possibly EAN. However, this violates most ToS.
2. **Zalando API (if available)** — Check if Zalando has a partner API that exposes GTIN/model data.
3. **Image-based matching** — Use the product images (which ARE provided) to match against known sneaker databases via visual similarity/AI.
4. **Brand + price + color heuristic** — With brand_name, colour, search_price, and Fashion:suitable_for, a fuzzy matching approach could narrow down candidates, but accuracy would be low.
5. **Accept the limitation** — Show Zalando Lounge products as "brand sneakers on sale" without specific model matching, and link to the product page where users can see the actual model.

### Bottom line:
No feed configuration change will fix this. Zalando Lounge simply doesn't provide product identifiers through Awin. The matching problem requires either an alternative data source or a different matching strategy (image-based or accept generic listings).